Marketing isn’t just about what you know; it’s also about who you know. T-Mobile uses social network analysis (SNA) to identify influencers within customer communities so it can focus its marketing dollars and, ultimately, decrease churn.
SNA is the study of customer interactions for the purpose of identifying communities and community influencers, says Eric Helmer, T-Mobile’s senior manager of campaign design and execution. T-Mobile classifies these communities, which average about 18 subscribers each, by looking at customers’ call records, such as their voice calls and text messages. This gives the telecom a starting population of about 200 million phone numbers. T-Mobile then excludes phone numbers that don’t have enough data points, such as those with low call or text volume or low reciprocity. These exclusions bring the population down to about 89 million phone numbers—23 million T-Mobile phone numbers and 66 million off-network numbers.
Using SAS’s Customer Link Analytics software, T-Mobile visualizes customer connections by drawing arrows between individuals, with the head of the arrow pointing from the call originator to the call receiver. The thickness of the arrow’s line also signals the strength of a connection—a thicker line indicates a stronger connection. The software then looks at approximately 200 SNA attributes, such as closeness, and condenses the attributes into four main scores: centrality, outbound connections, outbound usage, and connection to churn. Helmer says that centrality, or “how many connections come to you,” is the most influential factor because it has the greatest association with virality, which is the effect (or churn rate) influencers have on followers.
Each subscriber then receives an influencer score and T-Mobile determines a threshold to distinguish influencers from subscribers. According to Helmer, influencer churn increases follower churn by 25%.
“Friends and family are huge influencers on the decisions you make,” Helmer says, “and that’s what we’re able to tap into because people receive calls and texts from their friends and family.”
In 2012 T-Mobile ran an SNA test campaign in which it identified 15,000 influencers and 15,000 non-influencers and sent them an SMS message offering $50 off a handset upgrade. When influencers took the offer, the take rate among non-influencers nearly doubled, according to Helmer. By getting customers to upgrade, T-Mobile got customers to commit to another two-year contract. However, he says that this was before T-Mobile launched its Simple Choice plans that don’t require contracts; so, the brand will have to conduct another test to determine whether the model will still work.
T-Mobile faces three key challenges when using SNA modeling: ensuring that the brand stays within legal privacy rights and corporate policy, and that the brand cooperates efficiently across departments, Helmer says. For example, while T-Mobile’s engineering department holds the call records, the company’s marketing department needs the data to analyze and score the customers, he explains. However, T-Mobile has been able to overcome these challenges.
In fact, SNA modeling helps T-Mobile differentiate itself from its competitors, Helmer says. “We were able to optimize marketing spend based on this in-house Big Data,” he says, “and that’s a competitive advantage that some other industries don’t have.”